Eligibility interviewers

Three parts of your work where AI is already doing real lifting, and what stays yours.

• Paper case file + manual ledger1935–1970

  Paper case file + manual ledger (New Deal eligibility era)

  The eligibility interviewer of the New Deal era worked with paper: a hand-written application form, documentation the applicant brought (pay stubs, tax records, rent receipts), and a paper case file that lived in a physical cabinet in the local welfare office. Determinations were made by the worker based on printed program rules and supervisor review. The telephone was the main tool for verifying information with employers or other agencies. No computer, no database, no cross-program matching. The worker held essentially all program knowledge in their head and in the policy manual on their desk.

• Mainframe batch processing + Electronic…1970–1990

  Mainframe batch processing + Electronic Data Systems (EDS) welfare computing contracts

  By the early 1970s, EDS (Electronic Data Systems) and similar firms had begun winning contracts with state and federal welfare agencies to place benefit payment records, eligibility data, and case files on mainframe computer systems. In the District of Columbia and other large urban welfare systems, caseloads had grown too large to manage from paper files alone. Mainframe systems enabled batch payments -- the welfare check could be printed automatically for eligible recipients without a worker writing it by hand each month. For eligibility interviewers, this was a partial shift: initial determinations were still conducted in face-to-face interviews with paper forms, but the records were now entered by data-entry clerks into the central system after the interview. The interviewer was not yet directly interacting with the computer.

  Effect on the work

  Mainframe processing reduced clerical labor for payment processing but did not yet automate eligibility determination. The interviewer's core judgment work was unchanged; the data-entry and file-maintenance functions adjacent to the interview were the first to be streamlined.

• Desktop case management software + stat…1990–2005

  Desktop case management software + state integrated eligibility systems (IES)

  The early 1990s brought PC-based case management software into welfare offices. Human service agencies began investing in desktop systems (FamCare, Social Solutions ETO) that let eligibility interviewers directly enter case notes, eligibility determinations, and benefit amounts, bypassing the separate data-entry layer. States also began building Integrated Eligibility Systems that consolidated formerly separate program databases: a worker could now pull up a client's SNAP, Medicaid, and TANF records in one interface rather than checking three separate paper files. The interviewer's screen became the center of their work. By the late 1990s, most large state welfare agencies had at least a partial IES, and the face-to-face interview was now structured around the fields on the computer screen.

  Effect on the work

  Desktop systems reduced the time per case for routine determinations and enabled supervisory tracking of worker caseloads and error rates. Studies from the late 1990s suggested that IES adoption improved accuracy on simple cases but did not reduce overall staff needs because program complexity and caseload volume continued growing.

• Online applications + call-center eligi…2004–2015

  Online applications + call-center eligibility models (Indiana IBM, Texas TIERS, Maryland CARES)

  In the mid-2000s, several states moved aggressively toward digitizing welfare intake: Indiana contracted with IBM to replace face-to-face eligibility interviews with online applications and private call-center workers; Texas built TIERS, a rules-engine system intended to automate eligibility calculations; Maryland built CARES. The premise of each was the same: move applicants online, reduce the number of in-person eligibility interviewers, cut costs. Indiana's experiment became the most-studied failure in welfare technology history. After three years and over $500 million spent, the system had produced more than 700,000 erroneous benefit denials. Indiana terminated the IBM contract in 2009 and implemented a hybrid model combining online applications with human in-person caseworkers. The Indiana case established that automated systems could handle simple income-verification cases but failed systematically on complex households, multi-program cases, and applicants with disabilities or language barriers.

  Effect on the work

  The Indiana IBM failure and similar experiences in other states produced a lasting shift in policy consensus: online applications could reduce intake volume and first-contact labor, but human eligibility interviewers were still required for determination and for the cases that automated pre-screening flagged as complex. The ambition to fully automate the interviewer role was not realized.

• ACA Medicaid expansion + automated dete…2014–2023

  ACA Medicaid expansion + automated determination portals (HealthCare.gov, state exchanges)

  The Affordable Care Act's Medicaid expansion (effective January 1, 2014) extended eligibility to more low-income adults in participating states, adding millions of new Medicaid enrollees by 2015. The ACA also required states to build real-time eligibility determination systems that could check applicants' income against IRS and SSA data and render a determination within seconds for simple cases. This was the most significant technical modernization of the eligibility determination role to that point: routine income-verification cases for straightforward households were handled automatically, with the eligibility interviewer only touching the cases that the system flagged as requiring human judgment. In states that built effective systems, the simple-case intake load per interviewer declined while the complexity of the remaining cases increased.

  Effect on the work

  By 2023, several states reported automating a large share of Medicaid and CHIP eligibility determinations at the point of application submission. The automation share was concentrated in simple household types; complex cases (self-employed, mixed-status families, disability co-morbidities) remained predominantly human-processed.

• AI-assisted eligibility tools + Medicai…2023–present

  AI-assisted eligibility tools + Medicaid unwinding surge (Salesforce Agentforce, Appian AI)

  Two things converged in 2023 and 2024 to reshape the eligibility interviewer's work environment. First, the COVID-era continuous coverage requirement for Medicaid ended in April 2023, requiring states to conduct eligibility re-determinations for all enrollees (tens of millions of people) within several months. Tens of millions of people lost coverage during the process, many due to procedural issues rather than actual ineligibility. The workload surge overwhelmed eligibility teams across the country. Second, AI-assisted tools moved from pilot to production in several states: Salesforce Agentforce for Public Sector offered constituent intake automation, Appian Case Management Studio used LLMs for document extraction and case routing, and the USDA explicitly encouraged states to explore AI to address staffing challenges in SNAP. As of 2026, the consensus among policy experts is that AI tools reduce administrative burden on eligibility workers but do not replace them: the legal requirement for a human determination authority, the complexity of multi-program edge cases, and the client-service dimension of the role all remain human.

  Effect on the work

  The 2023-24 Medicaid unwinding established that automated eligibility systems could produce high rates of procedural errors when applied to large-scale re-determination events. The experience strengthened the policy consensus that AI tools should augment eligibility workers, not replace them: the Route Fifty coverage of the 2026 staffing debate found experts consistently recommending human-centered design of AI tools with eligibility workers' needs as a design constraint.